基于动态贝叶斯网络的建筑火灾蔓延多米诺事故风险评估方法研究OA

The spread of a building fire is a high-impact and low-frequency event,exhibiting domino-like characteristics,often leading to more severe consequences than the initial fire.To accurately describe the domino effect of fire spread in buildings,a risk assessment method based on dynamic Bayesian networks(DBNs)for domino accident scenarios in building fire spread is proposed.The theory of the domino effect is introduced to describe the fire propagation process.By decoupling fire spread,the chain-like escalation process of undesirable events that trigger secondary accidents is analyzed,and a method for calculating escalation probabilities is proposed.Rooms are treated as nodes,and fire spread paths are treated as directed edges to construct the DBN structure.Each node encompasses four states,that is,safe,ignited,fully developed fire,and decay,with escalation probabilities serving as input parameters for the conditional probability tables.The effectiveness of the method is validated through fire spread experiments in confined spaces,and its application to identifying high-risk rooms and simulating fire scenarios during an incident is demonstrated.Results indicate that the proposed method can effectively identify critical units and prioritize risks.By updating the DBN with real-time evidence,the rooms most susceptible to fire spread in future time steps can be determined.The accuracy of DBN inference depends on the precision of the conditional probability tables.However,even if the escalation probabilities are not sufficiently accurate,the dynamic Bayesian network model can still correct simulation results in real time as new evidence emerges,demonstrating good robustness.